Geochemical classification for bottled natural waters in China: using unsupervised and supervised machine leaching algorithm

Major elements in still bottled natural water were analyzed in China. An unsupervised machine learning algorithm, principal component analysis (PCA), was used to classify the samples. The PCA result suggested four groups to discriminant the chemical composition of the samples. By using the labelled data, supervised machine learning methods, random forest, and support vector classification, were used to train models. The models were then applied on the data obtained from literature. To analyze the water- rock interactions of the samples from different groups, reverse modeling in the software PHREEQC was implemented.